sift

所属分类:matlab编程
开发工具:matlab
文件大小:2765KB
下载次数:118
上传日期:2012-05-25 15:31:16
上 传 者fireguard
说明:  1 SIFT 发展历程 SIFT算法由D.G.Lowe 1999年提出,2004年完善总结。后来Y.Ke将其描述子部分用PCA代替直方图的方式,对其进行改进。 2 SIFT 主要思想 SIFT算法是一种提取局部特征的算法,在尺度空间寻找极值点,提取位置,尺度,旋转不变量。 3 SIFT算法的主要特点: a) SIFT特征是图像的局部特征,其对旋转、尺度缩放、亮度变化保持不变性,对视角变化、仿射变换、噪声也保持一定程度的稳定性。 b) 独特性(Distinctiveness)好,信息量丰富,适用于在海量特征数据库中进行快速、准确的匹配[23]。 c) 多量性,即使少数的几个物体也可以产生大量SIFT特征向量。 d) 高速性,经优化的SIFT匹配算法甚至可以达到实时的要求。 e) 可扩展性,可以很方便的与其他形式的特征向量进行联合。 4 SIFT算法步骤: 1) 检测尺度空间极值点 2) 精确定位极值点 3) 为每个关键点指定方向参数 4) 关键点描述子的生成 本包内容为sift算法matlab源码
(1 SIFT course of development SIFT algorithm by DGLowe in 1999, the perfect summary of 2004. Later Y.Ke its description of the sub-part of the histogram with PCA instead of its improvement. 2 the SIFT main idea The SIFT algorithm is an algorithm to extract local features in scale space to find the extreme point of the extraction location, scale, rotation invariant. 3 the main features of the SIFT algorithm: a) SIFT feature is the local characteristics of the image, zoom, rotate, scale, brightness change to maintain invariance, the perspective changes, affine transformation, the noise also maintain a certain degree of stability. b) unique (Distinctiveness), informative, and mass characteristics database for fast, accurate matching [23]. c) large amounts, even if a handful of objects can also produce a large number of SIFT feature vectors. d) high-speed and optimized SIFT matching algorithm can even achieve real-time requirements. e) The scalability can be very convenient fe)

文件列表:
sift\data\box.pgm (72310, 2008-04-17)
sift\data\box.sift (233055, 2008-04-17)
sift\data\circle.pgm (10060, 2008-04-17)
sift\data\circle.sift (1362, 2008-04-17)
sift\data\img3.jpg (464954, 2008-04-17)
sift\data\img5.jpg (467808, 2008-04-17)
sift\data\landscape-a.jpg (275111, 2008-04-17)
sift\data\landscape-b.jpg (304723, 2008-04-17)
sift\data\nest.png (11101, 2008-04-17)
sift\data\nest2.bmp (17462, 2008-04-17)
sift\data\vessel-1.pgm (337980, 2008-04-17)
sift\data\vessel-2.pgm (337980, 2008-04-17)
sift\diffss.m (2797, 2008-04-17)
sift\doc\bibliography.bib (246, 2008-04-17)
sift\doc\figures\sift-descriptor-AI.bb (550, 2008-04-17)
sift\doc\figures\sift-descriptor-AI.pdf (475918, 2008-04-17)
sift\doc\figures\sift-descriptor-SAVED.tex (3275, 2008-04-17)
sift\doc\figures\sift-descriptor.aux (8, 2008-04-17)
sift\doc\figures\sift-descriptor.log (13430, 2008-04-17)
sift\doc\figures\sift-descriptor.pdf (443880, 2008-04-17)
sift\doc\figures\sift-descriptor.tex (3275, 2008-04-17)
sift\doc\figures\sift-descriptor.wrm (13, 2008-04-17)
sift\doc\figures\warmread.sty (48557, 2008-04-17)
sift\doc\sift.bbl (239, 2008-04-17)
sift\doc\sift.tex (22461, 2008-04-17)
sift\doc\visionlab.sty (6174, 2008-04-17)
sift\gaussianss.m (7920, 2008-04-17)
sift\imreadbw.m (2359, 2008-04-17)
sift\imsmooth.c (3960, 2008-04-17)
sift\imsmooth.dll (8192, 2012-05-24)
sift\LICENSE (1990, 2008-04-17)
sift\Makefile (4345, 2008-04-17)
sift\mexutils.c (2178, 2008-04-17)
sift\plotmatches.m (10001, 2008-04-17)
sift\plotsiftdescriptor.m (5466, 2008-04-17)
sift\plotsiftframe.m (5222, 2008-04-17)
sift\plotss.m (2734, 2008-04-17)
sift\sift.m (10029, 2008-04-17)
sift\siftdescriptor.c (16452, 2008-04-17)
... ...

AN IMPLEMENTATION OF LOWE'S SCALE INVARIANT FEATURE TRANSFORM Andrea Vedaldi vedaldi@cs.ucla.edu http://www.cs.ucla.edu/~vedaldi/ This is a MATLAB/C implementation of the SIFT detector and descriptor [1]. See the documentation in 'doc/' to get started. COMPLING From MATLAB prompt > sift_compile From shell prompt (Linux, Mac OS X) > make TRYING OUT THE CODE The package includes a number of test function called sift_demo*.m. ABOUT THE SOURCE CODE We use the follwing convention to name files: *.m *.c: M and MEX files sift*mx.c: `Private' MEX files sift_*.m: Service scripts (compiling, demos) CHANGES 0.9.14 Fixed issues in sift_compile.m on Mac Intel platforms. 0.9.13 Improved make dist. 0.9.12 Removed LAPACK call. Added LOWE_BUG symbol definition to reproduce a small issue with Lowe's implementation in the computation of the keypoint orientations and match its output more closely. 0.9.11 Corrected a bug which prevented Magnif to function properly. Removed a spurious KEYBOARD command. 0.9.10 Added sift_demo6.m. 0.9.9 Added a flag to compile on Intel Macs. Improved the documentation. 0.9.8 Minor bug fixes 0.9.7 Fixed a minor glitch in the convolution code. Increased the default number of octaves that get computed (this is more similar to Lowe's version). 0.9.6 Modified the convolution code to handle image boundaries more similarly to Lowe's version. 0.9.5 Fixed a minor issues with the new sift_demo3.m. 0.9.4 Fixed another bug in the detection of keypoints close to the boundary. Reduced the size of the distribution. New test code. 0.9.3 Improved plotmatches.m and tightsubplot.m 0.9.2 Improved detection of keypoints too close to the boundary 0.9.1 Fixed a bug in the creation of the .zip archive 0.9 Fixed a bug for which some of the keypoint detected where dropped. CREDITS The SIFT algorithm [1] has been patented by David Lowe. Some of the images in 'data/' are from [2,3]. [1] D. G. Lowe, "Distinctive image features from scale-invariant keypoints," IJCV, vol. 2, no. 60, pp. 91 110, 2004. [2] K. Mikolajczyk, T. Tuytelaars, C. Schmid, A. Zisserman, J. Matas, F. Schaffalitzky, T. Kadir, and L. Van Gool, "A comparison of affine region detectors," IJCV, vol. 1, no. 60, pp. 63 86, 2004. [3] C. Hormann, "Landscape of the week 2," 2006.

近期下载者

相关文件


收藏者